Comment

Comment on Artists lose first copyright battle in the fight against AI-generated images

I have a feeling you may not have seen the difference between how digital artists are using generative AI in workflows and users just creating generative AI images from a prompt with no additional work.

The idea that digital art capabilities are going to disappear because diffusion models can generate digital art may be a bit too binary of a consideration from the reality.

You are right, that photography did replace a lot of still life work, such as in magazine ads, etc. But it only reduced the market for the skill set, and many people still produce ‘photorealistic’ art today.

I’d agree the market for drawing and especially prototyping is going to be made more efficient, but I’m skeptical it’s going to be entirely reduced as you put forward by your analogy.

Making something like a movie poster already went from making 10 mock-ups by hand from a team of artists to making a hundred mock-ups compositing using asset libraries and moving forward will likely end up in a place where there’s 1,000-10,000 versions which are each run though virtual focus groups to create a selection set for the client.

But the final product will absolutely still involve digital artists, and if anything the component that’s mostly being replaced is the asset library, along with around a 10x or more time savings on an individual artist’s generation.

That will either result in a 10x increase in variations or 1/10th the staffing or somewhere in between, but as mentioned parallel advances in AI mean that significantly increased output is very likely going to have significantly increased return, so you may even ultimately see slightly larger digital art teams from today as time goes on.

There’s a bizarre assumption that modern labor output represents a demand cap and as such efficiency in supply means less people making the same amount of things.

That’s almost never historically been the case during industrialization and unless the role becomes entirely obsolete, scaling up productivity with a new tool will bias towards increased output not decreased suppliers - outside of decreased demand for suppliers who have eschewed the new technology and efficiencies.

source

Sort:hotnew top

ParsnipWitch@feddit.de ⁨10⁩ ⁨months⁩ ago
You are only arguing about labour output and corporate work but I explicitly wrote about the impact on how many people in the future will invest in learning skills like digital illustration, drawing and painting.

As you probably know, it’s not a three year apprenticeship and then you have an artist.

People learn how to draw, which takes an enormous amount of effort and time and often money as well, because they want to visualise their ideas as convincing as possible. On their way of doing that communities like Deviant Art are created and filled with their work.

I for example will not continue digital illustration, especially not posting it somewhere online. My work is easily replaceable by AI. I did commissioned fan art mostly and sci-fi illustrations for collectors sometimes. But it was only a side gig, I am in one of the bigger groups of artists who barely can pay for their hobby with what they earn. That’s over now.

Why would I continue investing so much time and money in a skill that’s not appreciated or paid anymore? When a younger person asks me if they should invest the tens of thousands of hours necessary to become a good artist I will tell them to better learn something else.

Perhaps there will be an increase in traditional drawing skills learned. But that’s vastly more expensive. Most people don’t have the money and space to learn how to create oil paintings for example.

Will companies still higher some people who know about digital art? Probably, for a while. But that’s not what I was writing about.

source
- kromem@lemmy.world ⁨10⁩ ⁨months⁩ ago
  There haven’t been sword duels in a long time. And yet people still learn fencing (which tends to only cost money and not make it).
  
  The commercial market for clothing doesn’t require people to learn to hand knit, and yet that’s a skill that’s still pretty popular.
  
  If you think people are going to stop drawing as a hobby, I guess I just don’t really see it that way.
  
  You may stop drawing out of a concern that an AI can use your images to learn how to draw similar images (though do keep in mind your images in the training set is about the relative equivalent of spitting in the ocean).
  
  And that’s entirely your prerogative. It’s your skill and output to do with what you want.
  
  But there’s plenty of artists who will continue to produce art both professionally and others as a hobby that they enjoy, and many will continue to share it.
  
  And the idea that drawing as a skillset is going to disappear is ridiculous.
  
  Did computerized synthesizers cause people to stop learning to play musical instruments? Because one half of US households today have someone who plays, even though there are alternative technologies to create very similar end results without the same investment of skill and patience.
  
  source
  - ParsnipWitch@feddit.de ⁨10⁩ ⁨months⁩ ago
    You are building a strawman. I never said the skill of drawing will vanish or no one ever will learn how to draw anymore. This is what I wrote:
    
    They will also very likely make less people pursue drawing, (graphic) design and painting as a skill.
    
    And right now you have wrote nothing to dispute that. Quite the opposite, you seem to agree with it. You just disagree with the strawman that no one will draw by hand anymore.
    
    The dynamics are also very different depending on the specific activity. Making music and sports as a hobby have other perks that drawing doesn’t have. And not everybody who doodles from time to time is an artist (in the sense of someone honing the craft).
    
    To make the ideology more fitting you would have to ask: How many people will continue to learn how to fence if we had ultra fast learning, extremely competent and getting even better fencing AI robots? They copy the fencing styles of all people they see fencing. They will from now on be at most fencing tournaments, they will be at all local gyms, Olympia, can be booked as teachers and used in movies. Instead of fencing themselves your opponents can just put a fencing AI robot into the duel.
    
    Do you really think this wouldn’t have a significant impact on the motivation to pursue fencing as a skill?
    
    source
    kromem@lemmy.world ⁨10⁩ ⁨months⁩ ago
    Photoshop also made drawing less common of a skillset as I originally said. But you took issue with that comparison because in your follow-up suggesting photography as a better analogy you specified that in that case it made the skill “incredibly rare” - your words, not my strawman.
    
    And no, what you described about fencing is exactly what happened to the hobby of ‘chess.’ AI could learn from players’ games. Could then compete against them and beat them. Copy their styles. And you can just play against it instead of humans.
    
    Has AI existing which can beat any human player in chess made that a dead hobby? (Hint: chess is experiencing a huge boom right now).
    
    source
Fungah@lemmy.world ⁨10⁩ ⁨months⁩ ago
God DAMN that’s a well reasoned and written comment that demonstrates a lot of familiarity with the material.

I hope digg 2.0 never happens to reddit. Lemmy needs to stay like this forever.

Only thing id add is that for the moment any ai generated art needs quite a bit of human intervention for it to be exactly what the artists envisions. You’re argument about the lack of cap on demand for what AI can generate is a great point, because that would need to be the case for people in some industries to keep jobs as ai progresses. In image generation were already starting to see the prompts needed to generate what one wants be less cryptic and more like natural language, though we ain’t there yet. It’s moving quickly though.

I think a lot of the uncertainty lies in not knowing for sure where a lot of this tech will land. Will it be able to write engaging, novel, and new scripts / books, or even entire movies one day? Or is it always gooing to be the eloquent, stupid dumpster firebthay is chatgpt?

If the tech never becomes seamless, competent, and all around useful, the need for human intervention increases. If it does, it decreases. Which doesn’t mean there won’t be jobs dealing with creating the input and directing the output, but unless regulations cut off access to these tools to all but the richest individuals and companies those jobs will be handling a commodity that is essentially post scarcity.

Most jobs today focus either on selling things people have created and are all governed in some way by scarcity - of natural resources, of the time it takes to create software or art, supply chains, etc. If ai is good enough the human input needed is trivial or even nonexistent, and the output governed only by computing resources, efficiency of the code, etc. Both of these CAN be(aren’t always - the compute cost of AI has ready raised eyebrows and running dozens of enterprise-grade GPUs isn’t exactly gree) so trivial st scale that the driving force behind what’s crested is a demand for curation stronger than any weve seen in the world. The noise:signal ratio is going to get so bad, and it wouldn’t surprise me if one day thousands of novels better than any human being has ever written lie unread on magnetic tape .

source